To meet the growing capacity demand of fiber optic communication system, more channels and higher line-rates have to be considered in dense wavelength-division multiplexing (DWDM) systems. These systems suffer from many propagation impairments, such as amplified spontaneous-emission (ASE), four-wave mixing (FWM), cross phase modulation (XPM), and stimulated Raman scattering (SRS).
All these effects limit the transmission distance and optical regeneration is necessary to restore optical data signals. In an all-optical network, the need is even more acute as the optical signal may not only travel over variable distance but also go through unpredictable numbers of switching nodes.
In conventional opto-electronic repeater/regenerators, two complete WDM (wave-division multiplexing) terminal equipments in a back-to-back configuration are used. They are expensive since both O-E and E-O conversion are required, and will be finally bandwidth-limited. In the future high speed long-haul transmission systems or large-scale optical networks, all-optical regeneration, including 2R (reshaping and reamplification) and 3R (2R plus retiming) are key technologies to overcome the electronic bottleneck.
Many approaches have been proposed for optical 3R regeneration. The majority of them can be classified into the following categories: 1) cross-gain modulation (XGM) regenerators using semiconductor optical amplifiers (SOA) and/or distributed feedback (DFB) lasers; 2) interferometric regenerators using cross phase modulation (XPM) in SOAs or optical fiber; 3) cross-absorption modulation in electro-absorption modulators (EAM); 4) soliton-based regenerators using synchronous modulation; 5) noise suppression on bit 1's using gain saturation in SOAs or fiber parametric amplifiers; and 6) regenerators based on spectral broadening due to self-phase modulation (SPM); 7) high-order parametric processes.
Regeneration schemes based on gain dynamics of SOAs are limited in speed by the carrier recovery time and in ER by weak gain saturation. As a result, it is unlikely XGM or gain saturation in SOAs will function at 40 Gb/s and above. Other schemes and their potential problems with these approaches are hereafter discussed to place the invention of this application in the proper context. The review is focused on nonlinear optical gating.
Interferometric Regenerators
Interferometric regenerators based on XPM in SOAs can be of the Mach-Zehnder (MZ), Michelson or delay interferometer types exploiting similar phase dynamics as used in the terahertz optical asymmetric demuliplexer (TOAD). A MZ regenerator uses two symmetric SOAs. The input data to be regenerated is split into two equal parts, one is delayed with respect to the other. The retimed clock pulses are also fed equally into the two MZ arms. Because of the relatively delay, the phase difference experienced by the clock pulses in the two arms is roughly a series of rectangular pulse of width τ and height π. The transfer function of the interferometer is sin[(Φ1−Φ2)/2]. The speed of operation of this regenerator is limited only by the rise time, not by the carrier recovery time. As a result, regenerators based on XPM in SOAs have been demonstrated to operate as high as 80 Gb/s (and XPM based wavelength conversion up to 160 Gb/s). The other advantage is that the sensitivity of such regenerators is among the highest of all the optical regenerators due to strong nonlinearities in SOAs. The disadvantages of this type of regenerator are as follows. Because data to be regenerated in general will have long-term average power fluctuations and short-term pattern-dependent effects, the data need to be pre-processed so that the peak power of each bit is the same. Pre-processing in general only optimize operation of bit 1's of the original data to be regenerated. Therefore only the noise of bit 1's (non-inverting regenerator) or bit 0's (inverting regenerator) will be reduced by the regenerator depending on the initial phase delay of the interferometer.
Interferometric regenerators based on XPM in fibers are in general realized in the form of a nonlinear optical loop mirror (NOLM). It exploits the phase difference between the co-propagating and counter-propagating path. The NOLM has ultra high speed potential. It is not very practical because of the interferometric stability with long (˜km) length of the fiber loop, and as in the case of SOA based XPM regenerators, it requires preprocessing.
Cross-Absorption Regenerators
All-optical regenerator based on cross-absorption modulation in EAMs exploits saturation effects in EAMs. This regenerator consists of an EAM with two inputs: the data to be regenerated and a probe laser. At bit 1's, the input data saturates the absorption of the EAM, leaving it transparent to the probe laser. At bit 0's the EAM is still absorptive to the probe laser. In order to obtain (thresholding effects) improved ER, the EAM should be biased at a very lossy state and strong injection power is required. The average power required for the data signal is on the order of +17 to +19 dBm. The speed is limited by the carrier recombination speed, which is in general faster than the carrier recovery speed of lasers. Speeds up to 40 Gb/s have been demonstrated. The disadvantages of cross-absorption regenerators are: 1) it has a very low sensitivity (high input power) and 2) the speed is limited to about 40 Gb/s.
Soliton-Based Regenerators
Soliton-based regenerator exploits the robustness of soliton pulses under synchronous modulation. Synchronous modulation has been used to transmit solitons over unlimited distance. Soliton-based regenerator first converts the regular dispersion-managed RZ pulses to soliton pulses, which is optically filtered and then synchronously modulated by a recovered clock signal (synchronous modulation). The soliton pulses (bit 1's) that emerge from synchronous modulation is robust while noise pulses (bit 0's) will disperse. In addition, synchronous modulation reduces jitter in soliton pulses as the centers of the pulses attract towards transmission peaks at exactly the clock rate. This process can be repeated, each time resulting in a better ER and smaller jitter until reaching an ER and a jitter floor. Although synchronous modulation of multiple channels can be envisioned, RZ-to-soliton and soliton-to-RZ conversions have to be performed in separate channels since the conversion conditions for each channel are different. In addition synchronous modulation requires that each channel have the same clock rate. This is not generally satisfied because different WDM channels often come from different sources with independent clocks. Despite the attention this scheme has received, this approach is not cost effective with OEO regeneration when the cost of preprocessing (RZ to soliton) and post processing (soliton to RZ), which needs to be performed on a per channel basis, is factored in.
Regeneration Using Gain Saturation in Fiber Parametric Amplifiers
This involves regeneration using gain saturation in fiber parametric amplifier (FPA). The pump for the parametric amplifier is a CW laser (for 2R) or retimed clock pulse train (for 3R) and the data to be regenerated is used as the probe. Before input into the FPA, the probe is amplified so that bit 1's will saturate the pump. As a result only bit 1's can be reshaped. It should be noted that this gain saturation is due to pump depletion when the pump power is transferred to the probe. Compared to SOA's with saturated gain, this scheme can operate at high speeds. Inherently, this scheme is not very competitive because reshaping of the bit 1's comes at the expenses of reduced ER. As a result, negative power penalty cannot be achieved. Practically, when a signal needs to be regenerated, its ER is already low and the reshaped signal with a further reduction in ER would not be able to transmit any further in fiber.
Regeneration Using Self-Phase Modulation
In regeneration using self-phase modulation, the input pulses to be regenerated have a spectral width on the order of Δω0˜1/τ, where τ is the pulse width. Due to the effect of SPM, the spectral bandwidth of the pulses broadens to ΔωSPM=Δω0(2π/λ)n2IpL, where Ip is the pulse intensity (which can fluctuate), n2 is the nonlinear refractive, λ is the wavelength and L is the length of the nonlinear fiber. After SPM, the pulses pass through an optical filter whose center frequency, ωf, is shifted with respect to the input signal carrier frequency, ω0 as ωf=ω0+Δωshift. If the spectral broadening of the pulse is small enough so that ωSPM/2<Δωshift, the pulse is rejected by the filter. If the pulse intensity is high enough so that ωSPM/2≧Δωshift, a part of the SPM-broadened spectrum passes through the filter. This regeneration scheme on surface is quite attractive. It uses only passive component, can lead to ER improvement. However, there is a major problem in terms of retiming with this scheme. First, it does not allow a retiming mechanism. Second, the intensity fluctuations (noise and pattern effects) in the input data can lead to significant jitter up to ±10% τ. To solve the retiming issue, this scheme has been combined with synchronous modulation.
High-Order Parametric Processes
Recently optical regeneration using high-order parametric processes has been proposed. It relies on multiple nonlinear optical interactions that involve multiple pumps and multiple idlers. As such, it is complicated, requires complicated filtering, and has limited dynamic range.
Thus, the need exists for solutions to the above problems of the prior art.